Havard Eidnes writes:
> > Here is the disklabel for sd5; sd1-5 are all the same
> > ...
> > 7 partitions:
> > #        size  offset     fstype [fsize bsize cpg/sgs]
> >  a: 142264000       0       RAID                     # (Cyl.      0 -=
>   47997*)
> >  d: 142264000       0     unused      0     0        # (Cyl.      0 -=
>   47997*)
> >  g:     16002      63    unknown                     # (Cyl.      0*-=
>       5*)
> 
> If I'm not terribly mistaken, this spells trouble further down
> the road, even if you now did -I and -i.  The problem is that you
> allocate the entire physical disk to raidframe.  Where should
> then the disk label for the disk be stored?  I think the disk
> label needs to be stored outside of the area allocated to
> raidframe. 

RAIDframe doesn't touch anything in the first 16K of a "RAID" partition. 
A disklabel can live at the beginning of a partition without fear of 
being clobbered by anything RAIDframe does...

> Obviously this is at least partly machine-dependent,
> but the disk label is usually stored close to the start of the
> (NetBSD part of the) drive.
>
> What I think I would have done here is to start the RAID
> partition 63 sectors in, and would also probably remove the 'g'
> partition.
> 
> What's also odd and/or unconventional is that there is no 'c'
> partition.  The 'c' partition is used to indicate which part of
> the disk belongs to NetBSD (on i386, amd64) or indicate the raw
> partition (most if not all other ports).
> 
> Regards,
> 
> - H=E5vard

Later...

Greg Oster